Alternative to MIDI-CV

[gstopp at fibermux.com]

     Response to all the recent posts...
     
     It's true that a discrete 16-bit R-2R ladder has the LSB "down in the 
     mud", both from noise and component tolerance aspects. For MIDI-CV use 
     however you only really need 7 bits max per voltage. If you want the 
     14-bit resolution of pitch bend, then go ahead and use two 7-bit DACs 
     and sum the voltages together. This method has the added flexibility 
     of tweaking the pitch bend range as an analog attenuation, or "at the 
     receiver", as the MIDI spec puts it.
     
     For a 7-bit DAC, a discrete R-2R ladder is so easy that it's hard to 
     justify a DAC chip. The performance is totally acceptable for 
     equally-tempered pitch control. Just remember two things when 
     building- 1) drive the "rungs" of the ladder with well-defined 0's and 
     1's (like from CMOS inverters), and 2) get a big pile of 1% resistors 
     and a DVM and match them by hand if you want perfect intonation.
     
     As for my recent 16-bit R-2R ladder for the freq-voltage converter 
     experiment - it was just for kicks, really, as long as the bits were 
     there I thought I would throw them into the sum just to "sway the 
     noise a little". I then took Don Till's advice and tried the 
     exponential-discharge capacitor into a sample & hold idea. It looks 
     good on the scope, but I had trouble trying to get it to convert at 1 
     volt per octave using scope period measurement and a DVM. I probably 
     would have had better luck listening to a VCO while tweaking the 
     converter gain. Heck, as is it may be better than that goofy Korg 
     MS-20 pitch follower...
     
     Next I thought of an "exponentially discharging digital counter", in 
     which the 16-bit down-counter modifies its own clock in such a way 
     that the frequency division modulo is increased as the count 
     decreases. Now when the external input frequency transitions latch in 
     the period measurements, they will be at 1 volt per octave already (if 
     the counter modulo math is correct). Of course it's always possible to 
     have the period measurement value point to an address in a look-up 
     table, whose data value goes to the CV DAC. While this may seem 
     inelegant it will certainly work, but now we're getting into PROM 
     programming again. Geeez I guess I'm really latched on to this digital 
     approach (pun intended?) - seems to me there's more control over the 
     various parameters.
     
     Anyway this digital pitch recovery problem remains a background 
     project for me. It's a nice brain teaser, the kind that promotes that 
     "aha!" feeling when you least expect it. If anything comes up I'll 
     keep the list posted.
     
     - Gene
     gstopp at fibermux.com
     


______________________________ Reply Separator _________________________________
Subject: Re: Re[2]: Alternative to MIDI-CV
Author:  "J.D. McEachin" <jdm at synthcom.com> at ccrelayout
Date:    2/14/96 6:06 AM


Thinking about the best DAC to use for MIDI-CV...
     
Why not use a 16bit R-2R ladder?  Since linearity and clock rate aren't 
as crucial w/ CVs as for audio, couldn't you get away w/ an R-2R built w/ 
10k resistor networks?  Is there some reason R-2Rs don't work w/ 16 bits?
     
JDM